Alexa, Rana, Sheila, Kiwamu,
We locked the DRMI tonight for the first time. It could stay locked for more than 10 minutes.
(A prep: PRMI locking with REFL signals)
Before moving onto the DRMI locking, we wanted to lock the PRMI on the sideband without using the AS detector. We first locked the PRMI with the conventional sensors i.e. REFL9_I for PRCL and ASAIR_45Q for MICH respectively. By exciting and looking at transfer coefficients, we figured out
REFL45_Q = -12.5 x ASAIR_RF45_Q for MICH readout
So, we put 1/-12.5 = -0.08 in the input matrix and this worked.
Also, the large fluctuation we saw in REFL_A_RF45 yesterday (alog 13968) was visible today as well before we transitioned to REFL_RF45_Q for MICH. We checked if this behavior is also visible in REFLAIR_A_RF45. This was actually visible in REFLAIR_RF45 as well. This incidicates the fluctuation comes from some kind of common place in the REFL path or the ASAIR_A is imposing this fluctuation in the MICH through its feedback. Once we transitioned onto REFL_A, we saw both REFL_A and REFLAIR RF45 signals suppressed. On the other hand, as expected, ASAIR then started wandering. We should check some clipping or some obvious things on the ASAIR detector tomorrow.
(DRMI lock)
We aligned the SRC by locking SRY with ASAIR_RF45. Also we adjusted the demod phase of ASAIR_B_RF90 such that the signal is maximized in in-phase. At the beginning, we adjusted the demod phase such that the PRMI carrier resonance gave negative values. This was then flipped by 180 deg after we locked the DRMI because this signal stayed at a negative value when the DRMI was locked.
To lock the DRMI, we reduced the MICH gain by a factor of 5 according to Anamaria's DRMI document, but eventually we ended up with a gain reduction of only a factor of three which resulted in repeatable locking. We did not have to change the PRCL gain from the PRMI locking gains as expected. We guessed the SRCL gain from the same document. We fiddled with the control sign and gain for an hour or so and eventually it started locking.
We coarsely adjusted the PRM and SRM M2 stages to maintain the DRMI lock for a long time. The gains are right now based on some models/guesses. So we will revisit these values tomorrrow.
(SRC hopping ?)
We are not sure what is going on, but the SRC seems to jump back and forth between two modes. It looks as if the hopping is triggered by some slow misalignment. Depending on time, it happened as frequent as once per a couple of seconds. And sometimes it did not happen for some time like 20 seconds or so. This is visible in the dark port camera and AISAIR_RF90 which went back and forth between a high and low values. The low value went to a negative number for some reason. We need to investigate this issue more to figure out what is happening.
Sheila, Rana, Kiwamu, Alexa, Daniel
We had not been able to lock the PRMI all of today. During the day's usual ISI shaking we noticed that the AS beam had gone nearly of the camera and so spent awhile chasing trends, etc.
After not finding the cause of that, we noticed that the AS45 and REFL45 signals had dropped off to nothing around 3AM local time. As far as we can tell, no one was on site at the time.
Afer some hunting, we found a particularly horrible electronics setup for the 45 MHz modulation amplifier in the H1 CDS electronics room on the same rack as the 9 MHz Marconi generator:
There was an unsecured ZHL-1A amplifier on top of the rack, powered by a DC bench supply (of the lowest quality available), and connected by damaged Banana cables. The banana power cables are intermittent and turn off the amplifier power once in awhile. We could see the power supply current vary to any value between 0 and 400 mA as we brushed the cables - I wouldn't trust this amplifier any more.
Several people disavowed any knowledge of this setup (even after we found https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=11986). It contained a few low quality L-Com cables and several cheap BNC barrels. We must never use BNC connections and these kinds of unreliable adapters in our low noise RF setups. Only N and SMA connections are allowed.
We have relocked the PRMI after removing this setup. The modulation voltage for 45 MHz is now 4x lower. It would be best to make a higher quality RF amplifier setup underneath the PSL table as was done at LLO. We'll be using this weaker modulation for another few days, but we want the higher modulation depth by mid next week, when we get back to the 3f locking.
The attached screen shot shows the minute trends of the REFL and AS 45 over the last 48 hours. There is a dead time, which is when the modulation depth was zero, but suprisingly the peak to peak of the signals was not reduced by a factor of 12 dB as would be expected after the removal of the 6dB attenuator and the 19 dB amplifier.
Here's a 1-minute video of a PRMI sideband lock, from the AS port camera. Kiwamu, Rana, and Sheila will write more about this later in the evening.
Rana, Kiwamu,
In response to Sebastien's study of the BS motion, tonight we took a close look at the BS motion when the PRMI was locked and unlocked.
The conclusion is that:
it is the LSC feedback that has been shaking the angle of the BS below 100 mHz through some kind of L-to-A couplings.
(What to do)
We need to re-design better L-to-A decoupling filters to reduce the coupling. According to a transfer function measurement, the coupling coefficient was about 0.1 [rad at the bottom stage / meters at the M2 stage]. Because of this coupling, both pitch and yaw motion in the BS increased their RMS by a factor of three, mostly from components below 100 mHz. In order not to excite the angular motion more than they normally are, we roughly need another factor of three improvements in the existing L-to-A decoupling filters.
(Some plots)
The 1st plot show various spectra when the PRMI was locked on the sidebands. The upper plot of the 2nd attachment shows a comparison of the oplev signal with and without the PRMI locked. It is evident that the LSC feedback makes the spectra worse below 100 mHz. The rms increased up to 0.1 urad in rms. Also, the bottom plot of the same attachment shows the transfer function of the couplings from the M2 stage witness sensors to the bottom stage oplevs.
Yesterday by the end of the day the main path through to the pericope had been aligned on the X-arm table. This morning we started with a projection as described in https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=13975. As part of this we realigned the beam through the mask to optimize the shape on the screen.
Later today we aligned the power meter, and ISS photodiodes. The QPDs were checked and are stlil aligned. We calibrated the photodiodes and then started to calibrate the rotation stage for the output power to the CP. However we found a maximum throughput of only 2.5W. Clearly we are losing a lot of power somewhere.
The table has a second IR camera located on the optical table. Before the adjustment during the projection the beam on this looked quite uniform. After finding that we were losing power I looked at the image on this camera, which is taken using a 50% splitter mirror from the beam going to the CP. The image is attached, but appears to show that we are now not well aligned through the mask. We'll need to redo this alignment, but it is surprising that the image from the projection and the image on the table look so different. Possibly this is just the lower resolution of the camera used to image the projected beam.
Alastair Heptonstall, Greg Grabeel The projection turned out to be bad as well. Clipping had started from fairly far upstream and the only thing that had changed was the manual rotation stage being removed. The beam was steered back into alignment, but another issue came up. The beam was hitting low on the mask so the flipper mount was removed and replaced with a non-mobile but correct height mount until the flipper mount can be made to the right height.Here we can see just how bad the clipping was. ![]()
![]() |
|
---|---|
![]() |
![]() |
RGA is baking overnight -> The 1.5" gauge pair pump port valve is closed as well as the 10" gate valve -> Leaving PT346B off while unpumped (i.e. while the 10" gate valve and 1.5" valves are closed)
On the BSC-ISIs, things have been running differently at LHO and LLO for quite a while.
I took a look at a random chamber (ITMY) for both sites to see how we can make the best out of both configurations.
LLO ITMY Status:
Stage 1
- X, Y: 45mHz blend, FeedForward from HEPI, Sensor Correction
- Z: 90mHz blend, FeedForward from HEPI, Sensor Correction
- RX, RY: 250 mHz blend
- RZ: 200mHz blend
Stage 2
- X, Y: 250mHz Blend
- Z, RX, RY, RZ: Damping only
LHO ITMY Status:
Stage 1
- X, Y, Z: 45mHz blend
- Z: 90mHz blend
- RX, RY, RZ: 250 mHz blend
Stage 2
- All DOFs: 750mHz Blend
As we can see, the main difference is on Stage 2: most of the DOFs are turned off (damping only) at LLO. On the other hand, they have more features (FeedForward, Sensor Correction) installed on ST1 then at LHO.
By looking at the results, we see pretty similar results in X and Y. LLO seems to do a little better at 0.5Hz (better blend design, sensor correction). But, by looking at the rotational degrees of freedom, we can see that having a high blend on ST2 at LHO doesn't affect the low frequencies and help a little around 1Hz (we see that also in the optical lever).
In conclusion, turning on the sensor correction at LHO seems to be the next step. At LLO, it seems that turning ON the stage 2 loops with high blend filters wouldn't hurt.
Sebastien - it's great to see this kind of comparison and combining of the benefits of both sites - thank you!
The y-axis units on the optical lever plot - it's in radians, I assume?
Peter- Yes! Sorry about the confusion, I did all those plots with the same template without seeing my mistake (same thing goes for the rotational degrees of freedom on the other plots)
The commissioners have been complaining about too much motion in Yaw on the BS (almost 2microns peak to peak at 10 mHz (!!)).
We know we have some pick up between Z and RZ (see the various SEI logs from the last 6 months), so I switched from a ~100mHz blend in Z to a 750mHz. Relaxing the actuation in Z reduced the pick up seen in RZ.
I was afraid to degrade pitch with this change, but we improved yaw by a factor of ~5 without changing the pitch integrated RMS (it introduced a peak around 1Hz though).
I've tried to implement a 250mHz blend, but no improvement...
Thus, the configuration that seems to satisfy the commissioners is:
- Tbetter: X, Y, RX, RY
-Tcrappy: RZ
- T750: Z
I'll check what is the status of this "pick up" on the other chambers ASAP.
Yes, this was certainly better for PRMI stability.
We now need to work some more on the angular stability of the other loops. Peeking over Kiwamu's shoulder, I see that the RMS of the other DRMI mirrors is ~0.1 urad (for 0.01 < f < 0.5 Hz). We should tailor the ISI and HEPI loops to move this RMS down to lower frequencies (0.01-0.1 Hz) so that the WFS loops can handle it.
I think that this is the configuration that we want for now, one comment is that it would be worth plotting the oplev signals only up to 19Hz (or 10Hz or whatever frequency it is that cuts off that ugly peak), the 29Hz peak is room fans, I'm guessing that the 19Hz peak is an oplev pier resonance, then we could get an RMS that better represents the optic motion.
TACQ driver S1100025 was replaced with newly modded (L1200226-v2) S1000356 as SRM M3.
J. Kissel, E. Merilh I had Ed export his test results for one channel for this chassis from directly from the SR785 onto a floppy disk. I then plotted the results and compared against the model we have of this modified driver, which we use to design the compensation filters. See attached. The message -- the driver modifications for this particular channel look great. By comparing the results of the channels in the usual aLIGO CDS way (comparing the magnitude and phase values at certain frequencies against previous measurements recorded in the same way), the rest of the channels check out too. I've stuck the raw data, and the script that's used to plot it in the suspensions SVN repo, /ligo/svncommon/SusSVN/sus/trunk/electronicstesting/coildrvtf/lho/Triple_Acq/S1000356/*.TXT, and /ligo/svncommon/SusSVN/sus/trunk/electronicstesting/coildrvtf/lho/Triple_Acq/S1000356/plot_S1000356_tfs.m As a part of the standard aLIGO procedure, I've also - updated the e-traveler, indicating that is has been modified, and then subsequently installed as H1 SUS SRM M3. - attached the same .pdf I attach here to the corresponding DCC document, S1000356 (now at v5). (If not this, then at least the new test report should be attached.) - updated the DCC filecard for the H1-SUS-C7 rack, S1301874, to replace S1100025 with S1000356. This now completes all hardware modifications needed for ECR E1400369, Integration Issue 936, and work permit 4850.
Swapped out Top acquistion coil drivers in PR2, SR2 Old New S1100020 S1100040 S1100009 S1100028 S1100032 S1100012 S1100052 S1100003 Waiting on tests to finish to replace SRM.
Nothing special as these are generic Lvl1 controllers. They work. The attached pdf shows all the details of all controllers.
There are a few high frequency (~400hz) peaks in the simulated Open Loop that intrude into our -20dB space on a couple dofs, mainly X. If there is any ringing, these controllers may need to diverge from generic.
The Previous Lvl1 controllers are still loaded in the Lvl 2 & 3 Fake Banks, so we can return to these if you suspect these controllers.
I added a ROBUST_ISOLATED to the Guardian ISI_HAM4 to use the new Lvl1 controllers. The SEI Manager is paused. HPI & ISI are executing
The Output Sym filters are the only thing we can't revert as I refit the new TFs for these.
The controllers shown above use the 01_28 Blends,...even though the filters that somebody switched this platform to are the 250mHz blends.
Alastair, Greg, Patrick Alastair tried to use the TCSX rotation stage to change the power this morning. Nothing on the medm screen seemed to indicate that anything was happening. When I request a search for home, the 'Actual Cmd Position', 'Actual Velocity' and 'Drive Time (ms)' change on the 'Rotation Stage Readback' medm, but the 'Counter Actual Angle' does not.
Vern checked and says that there is no interlock plug installed.
The interlock bypass plug has appeared to fix the issue. We're not sure where what was in there before, or where it had gone, but thanks to Vern we have a good solution in place. Fingers crossed this solves the issues.
Here are the new interlock bypass caps that Vern rigged up for us. Seems to have solved the issues with the rotation stage as no more errors popped up while doing the projection work.
A. Sevingy, J. Kissel, A. Staley Continuing with the installation of modified triple acquisition drivers into all lower stages of Recycling Cavity HSTSs, as per ECR E1400369, Integration Issue 936, and work permit 4850. Aaron will post serial numbers when the final install of SRM is complete, however, anxious to use the new drivers, Alexa and I followed yesterday's procedure (see LHO aLOG 13958, we started at Step 2) to make the necessary software changes to support the new drivers on H1SUSPR2 and H1SUSSR2. The only difference was that, unlike PRM, the "twos" needed BOTH stages of COILOUTF filters modified, since they did not have a prior modified driver installed in M2. new safe.snaps, filter files, and top-level models have been committed to the userapps repo. As with yesterday's changes, no DAQ restart was required.
Same has been done for SRM now.
The attached video is of the AS port durring a DRMI lock. Towards the end, there are a few of the mode hopping events that Kiwamu described in his alog.
Also attached is a screen shot of various settings with DRMI locked.
We are setting the intent bit to undisturbed, since as far as we know there aren't any sseismic transfer functions starting tonight and we are leaveing DRMI locked on the sidebands.
When activity starts in the morning, it would be good if the operator could set the intent bit (on the ops screen) to commisioning. This way det char people know when we left the IFO alone
Set intent bit to 'Commissioning' as per Sheila's request.
just for bookkeeping purpose:
the DRMI locking trial was taken place from 5:00-ish to 6:14-ish UTC last night. The first lock was achieved at around 6:14:00 UTC followed by some frequent short locks for 10 minutes. After the first lock there were several good locks.